A shotgun metagenomics approach was applied on all three size fractions of four samples from Lakes Cisó and Banyoles C-III. Identical reads were removed using CD-HIT (Li and Godzik, 2006 (link)). Annotation of metagenomic reads was conducted through the JCVI prokaryotic annotation pipeline (Tanenbaum et al., 2010 (link)) using Uniref100, PFAM, TIGRfam and KEGG (Kyoto Encyclopedia of Genes and Genomes) Orthologs (KO) databases for taxonomic and functional annotation. JCVI Metagenomics reports (http://jcvi.org/metarep ) were used for analysis and comparative metagenomics (Goll et al., 2010 (link)). KO annotation was used for functional analysis and KO counts were normalized according to the length of the read and the length of the target gene (Sharon et al., 2009 ). The communities and functional profiles found in each size fraction were highly similar (Supplementary Figure S1 ) and, therefore, we pooled all reads after normalizing for sequencing depth for subsequent analyses, which allows for a better comparison of metagenomes.
The functional analyses focused on the three main biogeochemical cycles for this type of lakes, that is, carbon (C), nitrogen (N) and sulfur (S) cycling. The genetic potential of the microbial community was analyzed following the C, N, and S marker genes (KOs) as reported by Lauro et al. (2011 (link)) with a few modifications. We amended this previous rubric by adding the anaerobic carbon fixation carried out through the Calvin cycle by Chromatiaceae, and additional genes for polysulfide reduction, nitrate reduction and nitrite oxidation. In addition, the genes pyruvate:ferredoxin oxidoreductase (porA/B) were not considered as marker genes for fermentation as in Lauro et al. (2011) (link), because they are key genes in the reverse tricarboxylic acid cycle used for carbon fixation by Epsilonproteobacteria abundant in our study lakes (Campbell and Cary, 2004 (link); Takai et al., 2005 (link)). Because both sulfide oxidation and dissimilatory sulfate reduction pathways are mediated by the same set of genes (aprA, aprB and dsrA) but are found in different families of bacteria, we assigned metagenomic reads to each pathway according to phylogeny, that is, sulfate reduction for Firmicutes and Deltaproteobacteria reads, and sulfide oxidation for Alphaproteobacteria, Betaproteobacteria, Chlorobiaceae and Chromatiaceae. Finally, for the sulfur-oxidizing Epsilonproteobacteria of the order Campylobacterales we specifically searched for sox genes (coding for thiosulfate oxidation) not currently available in the KEGG database. Marker genes used in the present work are shown inSupplementary Table S1 . Hierarchical clustering and heatmap plots were generated with R (R Development Core Team, 2012 ) using the library ‘seriation'. Metagenomic data have been deposited at CAMERA (Sun et al., 2011 (link)) under accession number CAM_P_0001174.
The functional analyses focused on the three main biogeochemical cycles for this type of lakes, that is, carbon (C), nitrogen (N) and sulfur (S) cycling. The genetic potential of the microbial community was analyzed following the C, N, and S marker genes (KOs) as reported by Lauro et al. (2011 (link)) with a few modifications. We amended this previous rubric by adding the anaerobic carbon fixation carried out through the Calvin cycle by Chromatiaceae, and additional genes for polysulfide reduction, nitrate reduction and nitrite oxidation. In addition, the genes pyruvate:ferredoxin oxidoreductase (porA/B) were not considered as marker genes for fermentation as in Lauro et al. (2011) (link), because they are key genes in the reverse tricarboxylic acid cycle used for carbon fixation by Epsilonproteobacteria abundant in our study lakes (Campbell and Cary, 2004 (link); Takai et al., 2005 (link)). Because both sulfide oxidation and dissimilatory sulfate reduction pathways are mediated by the same set of genes (aprA, aprB and dsrA) but are found in different families of bacteria, we assigned metagenomic reads to each pathway according to phylogeny, that is, sulfate reduction for Firmicutes and Deltaproteobacteria reads, and sulfide oxidation for Alphaproteobacteria, Betaproteobacteria, Chlorobiaceae and Chromatiaceae. Finally, for the sulfur-oxidizing Epsilonproteobacteria of the order Campylobacterales we specifically searched for sox genes (coding for thiosulfate oxidation) not currently available in the KEGG database. Marker genes used in the present work are shown in